- Home
- Skills
- D Oit
- Do Novelist Ai
- Gemini Websearch
gemini-websearch_skill
- TypeScript
0
GitHub Stars
3
Bundled Files
2 months ago
Catalog Refreshed
4 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill d-oit/do-novelist-ai --skill gemini-websearch- .gitignore12 B
- search_analytics.json825 B
- SKILL.md6.6 KB
Overview
This skill performs advanced web searches using the Gemini CLI in headless mode with the restricted google_web_search tool. It adds intelligent caching, automated result validation, retry logic, and analytics to make programmatic web research faster and more reliable. Use it when you need current information, documentation, or repeatable search workflows.
How this skill works
The skill invokes the Gemini CLI with a tool restriction so only google_web_search runs, then parses the JSON output produced by the headless call. It checks a MD5-keyed cache (default 1-hour TTL) before running queries, scores results for quality and relevance, and logs analytics like latency and cache hits. Validation flags and retry parameters can enforce minimum quality and relevance thresholds.
When to use it
- Looking up current events, breaking news, or recently published docs
- Automating repeated or batch research tasks
- When you need programmatic, reproducible web searches with caching
- When you require basic quality validation and analytics on search results
- When you want to restrict tool usage to web search only
Best practices
- Prefix queries with "search " and keep queries focused and specific
- Enable caching for repeated queries; disable for one-off breaking news
- Set quality and relevance thresholds appropriate to your use case
- Use batch mode for large multi-query research and enable validation for critical tasks
- Restrict tools in .gemini/settings.json to ensure only google_web_search is used
- Monitor analytics and adjust cache TTL, min-quality, and retry settings based on patterns
Example use cases
- Single quick lookup: run a headless search for a library or API change and validate results
- Batch research: run hundreds of targeted queries from a file and export structured results
- Quality-gated research: require min-quality and retry on validation failures for critical findings
- Analytics-driven optimization: use logged latency and cache hit rates to tune queries and caching
- Breaking news: disable cache to force fresh web searches for time-sensitive topics
FAQ
Gemini CLI JSON output does not include citations or grounding metadata; validation relies on content quality metrics, completeness, and relevance scoring.
How do I force fresh results instead of cached ones?
Use the --no-cache option or set cache_enabled to false in the config to bypass the MD5 cache and always run a live Gemini search.